Jamcorder sold 2,500 units: why the hardware was the easy part

⏱️ Lectura: 10 min

Chip Weinberger sold more than 2,500 units of his automatic MIDI recorder, the Jamcorder, and reached a conclusion that contradicts the industry’s classic mantra: hardware is not so hard. He tells the story himself on his personal blog, where he explains that the real bottleneck was the software: 200,000 lines of code that took him three years to write, in a world still without AI coding assistants.

📑 En este artículo

1. TL;DR
2. What happened
3. Context: from a software career to building hardware
4. Technical details: the board, the assembly, and what he decided not to build
5. The real bottleneck: 200,000 lines of code
6. How to get started: Weinberger’s checklist for launching hardware
7. Impact and analysis: why hardware is not so hard if you simplify it
8. What’s next
9. Frequently Asked Questions
   1. What is the Jamcorder?
   2. How long did it take to assemble the first 500 Jamcorders?
   3. Why was the software harder than the hardware in this project?
   4. What gross margin does Weinberger recommend for a hardware product?
   5. Does this conclusion apply to any hardware product?
10. References

Weinberger came from a career in software and wanted to accomplish a double goal: build the automatic piano recorder he had always dreamed of owning, and finally get into hardware for the first time. He expected it to be the hardest part of the project. It wasn’t.

TL;DR

• Chip Weinberger sold more than 2,500 units of the Jamcorder, an automatic MIDI recorder for piano.
• He hand-assembled the first 500 Jamcorders in just 4 days, without needing a single design change.
• The PCB has only 25 unique components, almost all off-the-shelf except for the custom MIDI connectors.
• The software (firmware, app, and manufacturing tools) added up to about 200,000 lines of code and took more than 3 years, in a pre-LLM world.
• He recommends targeting a gross margin of at least 70% to absorb costs, defects, and returns.
• Final assembly uses a single screw per unit and an injection mold with no slides.
• Among his tips: work with Chinese assembly houses via Alibaba and avoid single-source components.

What happened

The Jamcorder is a device designed for one job: automatically recording everything someone plays on a piano, without the user having to press any button. Weinberger designed it first for himself, as the recorder he’d always wanted to have, and it ended up becoming a product that has already surpassed 2,500 units sold.

To fine-tune the manufacturing process, he assembled the first 500 units with his own hands. It took him 4 days. As he tells it, he didn’t have to make a single design change along the way: everything worked on the first try. He expected something to go wrong (a ruined production batch, component sourcing issues) and that moment never came, although the tariffs imposed during the Trump administration came close to complicating his plan.

Context: from a software career to building hardware

Before the Jamcorder, Weinberger worked exclusively in software. The phrase “hardware is hard” is almost dogma in the startup world: electronics, plastics, manufacturing, fulfillment, component shortages. He had all of that in his head as a list of unavoidable obstacles before diving into building his first physical product.

The surprise was that none of those classic obstacles showed up at the scale he expected. What did consume years of work was the software: firmware for the device, the companion app, and internal manufacturing tools, about 200,000 lines of code in total. That development took more than three years and, as Weinberger himself clarifies, happened in a world before AI coding assistants, without Claude Code or Copilot to speed up the work.

Technical details: the board, the assembly, and what he decided not to build

The Jamcorder board has 25 unique components. Almost all of them are off-the-shelf, catalog parts, except for the MIDI connectors, which are custom made to order. Mechanical assembly comes down to a single screw per unit, and the injection mold for the housing has a generous draft angle and uses no slides, which greatly simplifies production tooling.

The key to that simplicity was what Weinberger decided not to build. He cut low battery detection, ambient light detection, a physical power button, and even USB-C. Each of those pieces would have added components, extra firmware, and more points of failure.

Feature cut	Why it was cut	Result
Low battery detection	Added a component and extra firmware logic	Fewer parts, fewer points of failure
Ambient light detection	Not essential for recording MIDI	Simpler PCB, one less chip
Physical power button	Complicates the mold and assembly	Single-screw assembly
USB-C	Extra connector and its control circuit	Cheaper BOM and fewer field failures

Weinberger is clear that this simplicity was a design decision, not a coincidence. If the Jamcorder had been 10 times more complex, or if it had to compete in a market like smartwatches or cars, the story would have been very different.

The real bottleneck: 200,000 lines of code

While the hardware moved forward without a hitch, the software became the years-long project. Firmware, app, and manufacturing tools added up to about 200,000 lines of code, written over more than three years, many nights included.

For a development team today trying to estimate how much it will cost to maintain a healthy margin against those development costs, a simple calculator helps put the gross margin into concrete numbers:

function grossMargin(sellingPrice, unitCost) {
  return ((sellingPrice - unitCost) / sellingPrice) * 100;
}

// Simplified example inspired by the Jamcorder case
const margin = grossMargin(199, 55);
console.log(`Gross margin: ${margin.toFixed(1)}%`);
// -> Gross margin: 72.4%

With a selling price of 199 and a unit cost of 55, the gross margin comes out to around 72%, above the 70% floor Weinberger recommends to absorb unexpected production costs.

How to get started: Weinberger’s checklist for launching hardware

Weinberger closes his post with practical recommendations for anyone thinking about manufacturing a physical product. Summarized, they can be applied right now during a project’s planning stage:

• Simple BOM: keep the bill of materials short and avoid single-source components whenever possible.
• Assembly without complex calibration: every extra step in the production line is a point of failure and cost.
• Partners in China: Alibaba is still the starting point for finding reliable assembly houses and suppliers.
• Gross margin of 70% or more: the cushion that lets you survive defects, returns, and tariffs.
• Lean company: scaling hardware is slower than scaling software, so the team doesn’t need to grow at the same pace.
• Anti-counterfeiting strategy: define it from day one, not after the copies have already appeared.
• Final QA in-house and finished stock in a local warehouse, not just at the factory.
• Request samples before every production run, no exceptions.
• Step-by-step manufacturing guide, with photos, so any operator can follow it.
• Compact packaging: a product that’s dense in value simplifies logistics and storage.

To spot upfront which BOM parts depend on a single supplier (one of the risks Weinberger mentions most), a short script over a materials CSV can flag the blind spot right away:

#!/usr/bin/env bash
# scans bom.csv and flags components with a single supplier
awk -F',' 'NR>1 {count[$1]++} END {for (c in count) if (count[c]==1) print c, "single supplier"}' bom.csv

The script assumes a CSV with the component name in the first column and one row per quoted supplier. Any component that appears only once is a candidate for finding a second source before committing to a large production run.

The path Weinberger followed, from prototype to 2,500 units sold, can be summarized in a fairly linear decision flow:

flowchart TD
A["Working prototype"] --> B["Hand-assembled (500 units in 4 days)"]
B --> C["Zero design changes needed"]
C --> D["Production scaled with a partner in China"]
D --> E["2,500 units sold"]

💡 Tip: Weinberger recommends targeting a gross margin of at least 70% before launching production: it’s the cushion that absorbs tariffs, defects, and returns without eating into the project’s profitability.

Impact and analysis: why hardware is not so hard if you simplify it

Weinberger’s conclusion isn’t that manufacturing hardware is trivial in general. He himself clarifies that the Jamcorder is, on purpose, a simple product, and that 2,500 units isn’t a huge scale by most industry standards. Even so, he insists the hardware side would have been manageable regardless, even at higher volume.

The nuance matters: if the Jamcorder were 10 times more complex, or if it competed in low-margin, deeply established markets like smartwatches or the automotive sector, the equation would change completely. The real lesson is more like “hardware is as hard as you make it,” and much of the fear surrounding building hardware is overblown when the product’s scope stays contained.

⚠️ Watch out: Don’t underestimate counterfeiting. Weinberger insists on having an anti-counterfeiting strategy from day one, something many hardware founders ignore until it’s already too late to fix.

For development teams in Latin America coming from the software world and weighing whether to make the jump to a physical product, the Jamcorder case works as a counterweight to the usual fear: the real risks tend to lie in feature complexity, not in the fact that the product is physical.

What’s next

The title of Weinberger’s post itself, “part 1,” suggests he’ll keep publishing about what he learned selling 2,500 Jamcorders, probably with more detail on fulfillment, support, and the software side that cost him so much. For anyone taking on a similar hardware project today, there’s a key difference from Weinberger’s experience: he wrote his 200,000 lines of code in a pre-LLM world. A team starting today, with code assistants as part of the daily workflow, starts with an advantage he didn’t have during those three years of long nights.

📖 Summary on Telegram: View summary

Try it yourself: if you’re about to launch a physical product, build your BOM in a spreadsheet and run the single-supplier script before committing to the first production run.

Frequently Asked Questions

What is the Jamcorder?

It’s a device created by Chip Weinberger that automatically records everything played on a piano, with no manual intervention from the user.

How long did it take to assemble the first 500 Jamcorders?

4 days, hand-assembled by Weinberger himself, without needing a single design change in the process.

Why was the software harder than the hardware in this project?

Because it added up to about 200,000 lines of code across firmware, app, and manufacturing tools, a development effort of more than three years carried out in a pre-LLM world, with no AI coding assistants.

What gross margin does Weinberger recommend for a hardware product?

At least 70%, as a cushion against tariffs, production defects, and returns.

Does this conclusion apply to any hardware product?

Not necessarily. Weinberger clarifies that the Jamcorder is intentionally simple and that in more complex products, or in low-margin markets like automotive, hardware difficulty can indeed be real.

References

• Chip Weinberger, “Hardware is not so hard”: the original post where the Jamcorder founder details the manufacturing process and his recommendations.
• Wikipedia: Bill of materials: what a BOM is and why simplifying it reduces supply risk.
• Wikipedia: Contract manufacturer: context on how the external assembly houses Weinberger recommends using actually work.
• Wikipedia: Gross margin: definition of the gross margin metric Weinberger uses as a key metric for his business.

📱 Enjoying this content? Follow @programacion on Telegram for daily tech content in Spanish: quick summaries, fresh content every day. @programacion

Imagen destacada: Foto de Vishnu Mohanan en Unsplash